Cyanobacterial harmful algal blooms release toxins and form thick blanket layers on the water surface causing widespread problems, including serious threats to human health, water ecosystem, economics, and recreation. To identify the potential drivers for the bloom, there is a need for extensive observations of the water sources with bloom occurrences. However, the traditional methods for monitoring water sources, such as collection of point ground samples, have proven limited due to spatial and temporal variability of water resources, and the cost associated with collecting samples that accurately represent this variability. These limitations can be addressed through the use of high-frequency satellite data. In this study, we explored the use of Random Forest (RF), which is one of the widely used machine learning architectures, to evaluate the performance of Sentinel-3 OLCI (Ocean and Land Color Imager) images in predicting bloom proxies in the western region of Lake Erie. The sixteen available bands of Sentinel-3 images were used as the predictor variables, while four proxies of the cyanobacterial masses, including Chlorophyll-a, Microcystin, Phycocyanin, and Secchi-depth, were considered as response variables in the RF models, with one RF model per proxy. Each of the proxies comes with a unique set of traits that can help with bloom detection. Among four RF models, the model for Chlorophyll-a performed the best with R² = 0.55 and RMSE = 20.84 µg/L, while R² performance for the rest of the other proxies was less than 0.5. This is because Chlorophyll-a is the most dominant and optically active pigment in water, while Phycocyanin, which is a strong indicator of harmful bloom, is present in low concentrations. Additionally, Microcystin, responsible for bloom toxicity, has limited spectral sensitivity, and Secchi-depth could be influenced by various factors besides blooms, such as colored dissolved organic and inorganic matter. On further examining the relationship between the proxies, Microcystin and Secchi-depth were significantly correlated with Chlorophyll-a, which enhances the usefulness of Chlorophyll-a in accurately identifying the presence of algal blooms.

Harmful algal blooms (HABs) threaten lake ecosystems and public health. Early HAB detection is possible by monitoring chlorophyll-a (Chl-a) concentration. Ground-based Chl-a data have limited spatial and temporal coverage but can be geo-registered with temporally coincident satellite imagery to calibrate a remote sensing-based predictive model for regional mapping over time. When matching ground and satellite data, positional and temporal discrepancies are unavoidable due particularly to dynamic lake surfaces, thereby biasing the model calibration. This limitation has long been recognized but so far has not been addressed explicitly. To mitigate such effects of data mismatching, we proposed an Akaike Information Criterion (AIC)-like weighted regression algorithm that relies on an error-based heuristic to automatically favor “good” data points and downplay “bad” points. We evaluated the algorithm for estimating Chl-a over inland lakes in Ohio using Harmonized Landsat Sentinel-2. The AIC-like weighted regression estimates showed superior performance with an R² of 0.91 and an error variance (${\sigma }_E^2$) of 0.29 μg/L, outperforming linear regression (R² = 0.34, ${\sigma }_E^2$ = 2.34 μg/L) and random forest (R² = 0.82, ${\sigma }_E^2$ = 0.92 μg/L). We also noticed the poorest performance occurred in the spring due to low reflectance variation in clear water and low Chl-a concentration. Our weighted regression scheme is adaptive and generically applicable. Future studies may adopt our scheme to tackle other remote sensing estimation problems (e.g., terrestrial applications) for alleviating the adverse effects of geolocation errors and temporal discrepancies.

The western region of Lake Erie has been experiencing severe water-quality issues, mainly through the infestation of algal blooms, highlighting the urgent need for action. Understanding the drivers and the intricacies associated with algal bloom phenomena is important to develop effective water-quality remediation strategies. In this study, the influences of multiple bloom drivers were explored, together with Harmonized Landsat Sentinel-2 (HLS) images, using the datasets collected in Western Lake Erie from 2013 to 2022. Bloom drivers included a group of physicochemical and meteorological variables, and Chlorophyll-a (Chl-a) served as a proxy for algal blooms. Various combinations of these datasets were used as predictor variables for three machine learning models, including Support Vector Regression (SVR), Extreme Gradient Boosting (XGB), and Random Forest (RF). Each model is complemented with the SHapley Additive exPlanations (SHAP) model to understand the role of predictor variables in Chl-a estimation. A combination of physicochemical variables and optical spectral bands yielded the highest model performance (R² up to 0.76, RMSE as low as 8.04 µg/L). The models using only meteorological data and spectral bands performed poorly (R² < 0.40), indicating the limited standalone predictive power of meteorological variables. While satellite-only models achieved moderate performance (R² up to 0.48), they could still be useful for preliminary monitoring where field data are unavailable. Furthermore, all 20 variables did not substantially improve model performance over models with only spectral and physicochemical inputs. While SVR achieved the highest R² in individual runs, XGB provided the most stable and consistently strong performance across input configurations, which could be an important consideration for operational use. These findings are highly relevant for harmful algal bloom (HAB) monitoring, where Chl-a serves as a critical proxy. By clarifying the contribution of diverse variables to Chl-a prediction and identifying robust modeling approaches, this study provides actionable insights to support data-driven management decisions aimed at mitigating HAB impacts in freshwater systems.

Precipitation plays a key control in the water, energy, and carbon cycles, and it is also an important driving force for land surface modeling. This study provides an optimal least squares merging approach to merge precipitation data sets from multiple sources for an accurate daily precipitation estimate in Northeast China (NEC). Precipitation estimates from satellite-based IMERG and SM2RAIN-ASCAT, as well as reanalysis data from MERRA-2, were used in this study. The triple collocation (TC) approach was used to quantify the error uncertainties in each input data set, which are associated with the weights assigned to each data set in the merging procedure. The results revealed that IMERG provides a better consistency with the other two input data sets and thus was more relied on during the merging process. The accuracy of both SM2RAIN-ASCAT and MERRA-2 showed obvious spatio-temporal patterns due to their retrieval algorithms and resolution limits. The merged TC-based daily precipitation provides the highest correlation coefficient with ground-based measurements (R = 0.52), suggesting its capability to represent the temporal variation in daily precipitation. However, it largely overestimated the precipitation intensity in the summer, leading to a large positive bias.

Soil moisture is a key variable mediating hydrological and energy exchanges between the land surface and atmosphere, and plays an important role in hydrometeorology and disaster management. Although various in-situ observation networks have been established, their spatial representativeness is limited. In this study, high-resolution soil moisture was estimated over the Miho River watershed using Sentinel-1 and the Light Gradient Boosting Model (LightGBM). A time-series analysis between Sentinel-1 C-band backscatter coefficients and soil moisture showed that the vertical-vertical polarized backscatter (σVV) responded more sensitively to soil moisture changes than the vertical-horizontal polarized backscatter (σVH). The statistical evaluation of the LightGBM-based soil moisture estimation yielded a correlation coefficient of 0.722 and a root mean square error of 4.69%. SHapley Additive exPlanation analysis indicated that the day of the year had the highest importance, and σVV showed greater importance than σVH. The spatial distribution of soil moisture revealed high soil moisture (25-40%) in cropland and grassland areas immediately after rainfall, whereas forest areas exhibited more gradual variation owing to canopy interception and shading effects.

Ground-based Global Navigation Satellite System (GNSS) vertical displacements provide an independent measure of hydrological loading, but their ability to resolve terrestrial water storage (TWS) variations in arid inland basins remains uncertain. This study compared vertical displacements from 15 GNSS stations within and around the Heihe River Basin with loading displacements derived from satellite-based GRACE and model-based Noah TWS anomalies. After removing non-hydrological signals from GNSS series, correlations among the network-mean vertical displacement for GNSS-GRACE, GNSS-Noah, and GRACE-Noah were 0.48, 0.49, and 0.74, respectively. Mean station-level correlations for the same comparisons were lower (0.38, 0.35, and 0.60, respectively), with weaker agreement within the basin interior. Time-series decomposition showed that GNSS captured inter-annual TWS variability more consistently than the seasonal cycle of TWS. GNSS generally peaks earlier than GRACE in the south but later in the north. Noah interannual variations diverged from both GNSS and GRACE, suggesting limitations in the model’s representation of arid-region hydrological processes. Overall, GNSS provides an independent constraint on long-term TWS variability and a diagnostic of land surface model performance. These findings support multisource monitoring of arid-basin water storage while highlighting the need for improved scale matching and in situ validation.